We provide evidence that mobile phone records can be used to predict the socioeconomic status and other welfare indicators of individual mobile phone subscribers. Combing several terabytes of anonymized transactional mobile phone records with data collected through 2,200 phone-based interviews, we test the extent to which it is possible to predict an individual's responses to survey questions based on phone records alone. We observe signicant correlations between asset ownership and a rich set of measures derived from the phone data that capture phone use, social network structure,and mobility.

Simple classication methods are able to predict, with varying degrees of accuracy, whether the respondent owns assets such as radios and televisions, as well as xed house- hold characteristics such as access to plumbing and elec- tricity. More modest results are obtained when attempting to predict a broader set of development indicators such as an individual's response to the question, "Have you had to pay unexpected medical bills in the past 12 months?" While these methods oer a powerful opportunity for policymakers and researchers working in developing countries, we argue that considerable calibration and renement is needed before such methods can be deployed.

Mobile phone ubiquity in much of the developing world has turned from a question of when rather than if. Some of the poorest and most remote parts of the world are being connected to the global telecommunications network to enable an unprecedented ability to both observe and interact with previously hard-to-reach populations at scale. While many mobile phone owners adopt this technology for basic phone use, the connectedness this expansive ownership enables presents an opportunity to the study and practice of economic development that extend beyond simple peer-to-peer communication.

The modern information technology sector and its underlying network infrastructure presented this same opportunity during its own formation. The network was not only valuable for the communication it enabled, but also for the data it produced from those who utilized its services. It also serves as a platform for a deluge of information systems and services that have become a part of our everyday lives and has spurred significant economic growth over the past few decades. This "data revolution" is well underway in the developed economies but is diminishing in its returns, solving increasingly marginal problems. This same transformation is relatively nascent in developing economies where more salient challenges, such as poverty, have yet to be overcome. In this dissertation, we explore a data-driven approach that leverages mobile phone technology to better measure and address poverty in sub-Saharan Africa.

Our approach starts with the identification of a problem: in this case, poverty. In the first chapter, we apply novel machine learning methods to analyze roughly ten terabytes of data of mobile phone use from Rwanda's largest telecommunications operator to measure poverty at a national scale. We demonstrate that an individual's history of mobile phone usage can be used to infer his or her socioeconomic status. Using this individual model of mobile phone use and socioeconomic status, we can predict poverty and wealth across the entire network and accurately reconstruct national and regional distributions of wealth. Once we obtain this measure of poverty, we can then focus our efforts in regions that are most afflicted.

The second chapter helps moves us from diagnosis to a potential cure. Predictions may be helpful to provide some guidance on which regions or populations to target but does not provide much in the way of what to do to have impact. In three years of field research in poor regions of rural Kenya and Rwanda, it was clear that much of the world's poor thrive and survive on subsistence agriculture, but many of these farmers also own mobile phones. Having such a platform enabled the ability to provide potentially welfare-improving information at scale. This chapter presents the research design and analyzes the results of of six randomized controlled trials testing the welfare effects of sending hundreds of text message formulations encouraging agricultural experimentation to over 500,000 farmers in Kenya and Rwanda. Targeting farmers with the right messaging and delivery characteristics was a focus of these trials. We find statistically significant effects on agricultural technology adoption and high rates of return on welfare outcomes by providing information over this medium. This mirrors the digital advertising industry in many developed economies and reminds us that advertisements as information can have very large welfare effects in poor information environments.

The third chapter dives deeper into one of the six studies where the research design focused on information spillover in Rwanda where mobile phone ownership was about half of what it was in Kenya. We find that information does indeed spillover onto other farmers within the same group, and those farmers who don't have phones experience the largest percentage increases in adoptions when others within the same group receive a text message. This has large implications on the effectiveness and cost efficiency of information treatments to regions with lower mobile phone adoption. Not only were these interventions effective, they were also very inexpensive and resulted in network effects, further improving agricultural technology adoption, increasing food production and reducing poverty.

The chapters in this dissertation develop a theory and methods for understanding how to leverage mobile technologies to measure and reduce poverty. It serves as a guide for both research and practitioners to approach solving problems in development that is grounded in measurement, data, collaboration, impact and scale.

Since its inception in 2001, Wikipedia has fast become one of the Internet's most dominant sources of information. Dubbed "the free encyclopedia", Wikipedia contains millions of articles that are written, edited, and maintained by volunteers. Due in part to the open, collaborative process by which content is generated, many have questioned the reliability of these articles. The high variance in quality between articles is a potential source of confusion that likely leaves many visitors unable to distinguish between good articles and bad. In this work, we describe how a very simple metric – word count – can be used to as a proxy for article quality, and discuss the implications of this result for Wikipedia in particular, and quality assessment in general.

Migrants play a central role in the economy and society of most developing countries and are primary drivers of economic mobility among poor and rural households. The decision to migrate is one of the most important economic decisions an individual can make. On the one hand, social networks play a crucial role in influencing people's migration decision. On the other hand, as migrants adapt to a new environment, their social network evolves. My research seeks to shed light on the influence of social networks on migration, as well as the influence of migration on social networks. This dissertation answers three questions on migration and social networks using large-scale social network data: (1) What are the roles of migrants in connecting global social networks? (2) How do social networks affect people's decision to migrate? (3) How do migrants' social networks evolve over the migration process?

In the first chapter, I explore in detail the relationship between international social ties and global migration. Social ties form the bedrock of the global economy and international political order. Yet prior empirical studies have been constrained by a lack of granular data on the interconnections between individuals. In this study, using several billion domestic and international Facebook friendships, I find that long-term migration accounts for roughly 83% of international ties on Facebook. By computing the average shortest path length in a social graph with and without migrants, I find that migrants effectively decrease the length of the average shortest path, and act as conduits for more shortest paths than non-migrants.

The second chapter studies how social networks influence an individual’s decision to migrate. Two distinct mechanisms through which social networks provide utility to migrants are disambiguated: first, that networks provide migrants with access to information, for instance about jobs and conditions in the destination; and second, that networks act as a safety net for migrants by providing material or social support. I use a massive `digital trace' dataset to link the migration decisions of millions of individuals to the topological structure of their social networks. The main analysis indicates that the average migrant derives more utility from `interconnected’ networks that provide social support than from `extensive’ networks that efficiently transmit information.

In the third chapter, I develop and validate a novel and general approach to detecting migration events in trace data. The most common `frequency-based' approach to inferring migration events often results in mis-classifications. The novel approach accurately classifies migrations, and also provides more granular insight into migration spells and types than what are captured in standard survey instruments.

The fourth chapter examines how migrants' social networks change over the migration and settlement process based on the migration events and dates that were detected in the third chapter. I characterize changes in network structure before and after migration by observing the evolving social networks of a nation's worth of migrants. I find stark and systematic changes in this structure: within two months of migrating, migrants cease communication with nearly half of their former contacts in their place of origin; these `lost' relationships are almost exactly offset by the 55% increase in new connections with people in the destination. I also show that friendship persistence and loss is highly predictable: the social ties most likely to persist are those that have frequent communication.

As a whole, the chapters in this dissertation develop methods and theories to understand the interaction between migration and social networks. It lays the groundwork for future researchers answering questions in migration and social networks using population-scale digital trace data.

Machine learning algorithms have started having an unprecedented impact on human society due to their improved accuracy. The ability to collect and analyze information at a large scale has enabled the researchers to develop novel algorithms that can beat the best of the human experts in quite a few cases. The size of the data and the quality of the data has been the primary factor behind the success of the machine learning algorithms. However, when it comes to the data related to human behavior a digital divide still exists. An indirect consequence of the popularity of the social networks and the proliferation of sensors in the developed world is that the researchers in industry and academia have been able to fine tune their findings and algorithms using these huge behavioral datasets providing accurate and deep insights about human behavior in the developed world. However, the same is not true about the developing world where until recently the surveys have been the primary way of collecting information about individuals in the society.

Social networks and digital sensors have not been that common in the developing world as compared to the developed world with one big exception, i.e., the mobile phones. More than 95\% of the world population today has mobile phone coverage and event in some of the most under-developed places of the earth the penetration of mobile phones is much higher as compared to other measures of human development like literacy or access to the financial infrastructure. As a result, researchers have been using the meta-data collected by the mobile phone companies in these developing countries as an alternative to the more conventional data sources. However, the raw mobile phone data may not be very well suited for the machine learning algorithms. In other words, there is a need for algorithms to convert the raw mobile communication meta-data into features suited for the machine learning algorithms. Developing novel ways to extract features from the mobile phone meta-data has been the central question of my research.

In this dissertation, I am going to describe my work on extracting features from mobile communication logs using techniques like Deterministic Finite Automata (DFA). I will also show that how this approach outperforms other methods for problems like product adoption and churn prediction. I further show that by using DFA based features and spectral analysis of the multi-view nature of mobile communication networks, advanced neural network training algorithms can be developed that beat the current state of the art methods for the problems like poverty prediction and gender prediction. Last part of this dissertation describes the value of communication networks data for research questions related to social networks analysis like what are the salient differences between the behavioral patterns of men and women in the developing world as exhibited in the communication networks data.